Online forums have become essential resources for information sharing and collaboration, but they also face the challenge of negative and offensive user feedback. To maintain a polite and safe online environment, an effective comment toxicity model must be developed. This project creates a model of comment toxicity using Deep Learning (DL) and Python. Using a DL architecture, the suggested model evaluates text data and classifies comments into non-hazardous and harmful categories. The model leverages powerful Natural Language Processing (NLP) methods, including word embeddings and recurrent neural network (RNN), to extract the semantic and contextual information from comments. Additionally, attention processes and CNN (CNN) are employed to enhance the model's performance. Several key components of the process include data preparation, feature engineering, model creation and training, and assessment. Python is a prominent programming language in the data science industry that is used to construct the workflow. Two open-source libraries that provide the resources needed to efficiently build and train DL models are TensorFlow and Kera’s. To evaluate the effectiveness of the model, a large dataset of tagged comments is used, which includes both hazardous and non-toxic remark occurrences. Recall, accuracy, precision, and F1-score are just a few of the assessment metrics used to gauge the model's performance. The project's output will improve content moderation systems by enabling platforms to recognize and report offensive comments as soon as they are posted. This work provides a practical and scalable approach DL-based comment toxicity detection, assisting online communities in fostering a more inviting and safer environment for its members.